cancel
Showing results for 
Search instead for 
Did you mean: 
intapiuser
Community Team Member
Community Team Member
Let's say you have a table that contains a row of expected values and a row of predicted values, and we want to determine the accuracy. What we need to do is find cases where the expected values are equal to the predicted values, count them, and divide them by the total number of rows.
 
Here's the SQL used to generate that. Note that I only want to find the accuracy on my test data (this dataframe was generated using a K-nearest-neighbors machine learning model. Check out this community post here if you'd like to learn more about how to build this dataframe), hence the where clause. 
 
In the below SQL, ESTIMATED_TARGET is the predicted data, ACTUAL_TARGET is the expected result
select
  sum(
    case
      when ESTIMATED_TARGET = ACTUAL_TARGET
        then 1
      else 0
    end
  )
  * 1.0 / count(*) as accuracy
from
  [knn_model]
where
  DATASET = 'test'
You can leverage a similar calculation if you're trying to determine what fraction of your rows meet a certain criteria!
Version history
Last update:
‎02-09-2024 11:13 AM
Updated by:
Community Toolbox

Recommended quick links to assist you in optimizing your community experience:

Developers Group:

Product Feedback Forum:

Need additional support?:

Submit a Support Request

The Legal Stuff

Have a question about the Sisense Community?

Email [email protected]

Share this page: